TVStoryGen: A Dataset for Generating Stories with Character Descriptions

TL;DR

TVStoryGen is a new dataset of professionally written TV episode recaps with complex character interactions, designed to advance story generation models that incorporate detailed character descriptions and evaluate faithfulness.

Contribution

The paper introduces TVStoryGen, a large-scale dataset for story generation with character details, and proposes reverse models for assessing story faithfulness, highlighting its potential for future research.

Findings

Hierarchical models with oracle content selectors perform best on automatic metrics.

Generated stories sometimes lack faithfulness to the brief summaries.

The dataset enables research on constrained story generation.

Abstract

We introduce TVStoryGen, a story generation dataset that requires generating detailed TV show episode recaps from a brief summary and a set of documents describing the characters involved. Unlike other story generation datasets, TVStoryGen contains stories that are authored by professional screen-writers and that feature complex interactions among multiple characters. Generating stories in TVStoryGen requires drawing relevant information from the lengthy provided documents about characters based on the brief summary. In addition, we propose to train reverse models on our dataset for evaluating the faithfulness of generated stories. We create TVStoryGen from fan-contributed websites, which allows us to collect 26k episode recaps with 1868.7 tokens on average. Empirically, we take a hierarchical story generation approach and find that the neural model that uses oracle content selectors for character descriptions demonstrates the best performance on automatic metrics, showing the potential of our dataset to inspire future research on story generation with constraints. Qualitative analysis shows that the best-performing model sometimes generates content that is unfaithful to the short summaries, suggesting promising directions for future work.